Generation of Stereoscopic Interactive Learning Objects True to the

Original Object

Diogo Roberto Olsen, Fl

avio de Almeida e Silva, Lucas Murbach Pierin, Aramis Hornung Moraes

and Edson Jos

e Rodrigues Justino

Programa de P

os-graduac¸

ao em Inform

atica - PPGIa, Pontif

ıcia Universidade Cat

olica do Paran

a - PUCPR,

Keywords:

Learning Objects, Stereoscopy, Immersive Environments, Virtual Reality, Multidisciplinarity.

Abstract:

Learning objects are used in many knowledge areas and may be aligned with technologies such as stereoscopy

and ultra-high deﬁnition, instigating interactions and arousing interest in educational environments. Gener-

ating such objects from real pieces is a challenge because it requires computational resources to maintain

quality and ﬁdelity. Another challenge is to port these objects to different devices such as immersive theater,

cell phones, virtual reality glasses, and televisions/projectors. Hence, we developed a framework capable of

generating learning objects from real pieces with quality and ﬁdelity of form, color, and texture. This article

focuses on the generation of learning objects for these devices.

1 INTRODUCTION

In a teaching-learning process, the teacher can deploy

elements that aim to arouse the students’ interest re-

gardless of their nature such as rocks, leaves, animals,

videos, images, etc. These elements enrich the pro-

cess by putting the students in contact with objects

that bring learning closer to their reality.

However, not all of these elements are easily ac-

cessible to the teacher, although representations may

be used to introduce them to the learners. Some of

these elements may be seen as Learning Objects (LO),

which, according to the Learning Technology Stan-

dards Committee (LTSC-IEEE), ”Learning Objects

are deﬁned here as any entity, digital or non-digital,

which can be used, re-used or referenced during tech-

nology supported learning”

. However, there is no

consensus on this deﬁnition, as seen in (Wiley, 2002).

Interactivity, high resolution, stereoscopy, aug-

mented reality and three-dimensional projection are

present in movie theaters, advertising, televisions and

cell phones. These techniques are available and may

be used to support the teaching-learning process as

a way of presenting and manipulating LOs. LOs

need environments and repositories to be shared, as

discussed in (Santiago and Raabe, 2010; Freire and

Fern

andez-Manj

on, 2016). There are LOs (Sinclair

LTSC - http://grouper.ieee.org/groups/ltsc/wg12/

et al., 2013) that can be used multidisciplinarily, some

can be customized(M

endez et al., 2016; Garrido and

Onaindia, 2013). LOs can be evaluated by indicators

themselves (Sanz-Rodriguez et al., 2010).

In order to create LOs compatible with these tech-

niques, we developed a framework composed of a

Full Frame Semi-spherical Scanner(F2S2) and pro-

cessing software. The F2S2 digitizes, through pho-

tography, all the angles of an object, forming a

hemisphere with these images; hence the term semi-

spherical (Section 2). The terms full frame refer to

the fact that each image represents the whole object

for that viewing angle, unlike other technologies, such

as the laser scanner that digitizes the object into parts

to later reconstruct a given view. The LO consists

of photos and navigation software, which allows for

interactive visualization, i.e., you can visualize and

manipulate the object from any desired angle. You

can associate points of interest with hypermedia. For

example, in a History class, one could pick up an ar-

chaeological object that, by its rarity, cannot be ma-

nipulated by the students. This piece can be scanned

for Ultra High Deﬁnition (UHD) presentation, allow-

ing for magniﬁcation and movement, showing details

that even the naked eye would not see. Explanatory

content (hypermedia) may be added at the teacher’s

points of interest to enrich knowledge. Figure 1 shows

an example of a LO in UHD and with hypermedia.

Characteristic for LOs generated by F2S2 is that

Roberto Olsen, D., de Almeida e Silva, F., Murbach Pierin, L., Hornung Moraes, A. and José Rodrigues Justino, E.

Generation of Stereoscopic Interactive Learning Objects True to the Original Object.

DOI: 10.5220/0006701402590266

In Proceedings of the 10th International Conference on Computer Supported Education (CSEDU 2018), pages 259-266

ISBN: 978-989-758-291-2

259

(a) (b)

Figure 1: LO: (a) Original, (b) with Hypermedia content.

they can be acquired and reproduced in UHD, which,

as mentioned, enables enlargement without loss of

image quality, allowing the visualization of details

as well as maintaining the ﬁdelity of shape, color,

and texture of the original object. LOs may also be

stereoscopic, allowing visualization with the notion

of depth. The stereoscopy is formed with stereoscopic

lenses or by the difference of angles in the photos ob-

tained by the F2S2, as will be seen in section 2.

Interactivity, as seen in (Koukopoulos and

Koukopoulos, 2017; Hung et al., 2017), allows the

object to be manipulated in three ways: scaling (as al-

ready mentioned), rotation and translation. Rotation

allows you to change the viewing point to any desired

angle. Translation allows a displacement of the image

within the projection, for example, horizontal, verti-

cal, and diagonal. Figure 2 shows the three compo-

nents of interactivity. In this interactive process, it is

possible to select points of interest and add a presen-

tation of contents, as seen in Figure 1.

(a) (b) (c)

Figure 2: Interactivity components: (a) rotation, (b) trans-

lation, and (c) scaling.

The content may be shown in different environ-

ments, such as a television (TV) or multimedia pro-

jector, which are common access in schools and uni-

versities and bring the beneﬁt of being stereoscopic

and/or 4K, and UHD, respectively. The characteris-

tics of these environments allow visualization with the

original quality. In addition, interactivity may be pro-

vided using input devices. Figure 3 shows an example

of an LO shown on TV.

An environment that is currently gaining promi-

nence, especially in advertising, is the Holographic

Pyramid (Sidharta et al., 2007; Rossi, 2015). Its oper-

ation consists of a monitor/TV/projector that projects

the image on a screen, usually made of glass, causing

Figure 3: LO on TV.

the sensation of three-dimensionality. The name pyra-

mid does not ﬁt all forms of this environment because

in some cases it shows only one face for visualization,

in others, there are three or four faces. Figure 4 shows

an LO in the holographic pyramid.

(a) (b)

Figure 4: LO in a Holographic Pyramid: (a) Lateral Projec-

tion, (b) Pyramid.

For teaching-learning purposes using the objects

generated by F2S2, the pyramid can be interactive, as

is the case of the TV. Besides presenting tridimension-

ality, each face of the pyramid offers a different point

of view.

Virtual reality (VR) glasses and mobile phones

(Hassan et al., 2013) may also be used in teaching-

learning since they are naturally stereoscopic as they

project an image in each eye. An advantage of RV

glasses are features such as the accelerometer and the

gyroscope, which allow the user’s movement to be

identiﬁed and applied to the LO, giving a spatial no-

tion. Figure 5 shows the RV Glasses and an LO on a

mobile phone.

Figure 5: LO on VR glasses.

Another environment that offers interactive and

stereoscopic features is the immersive theater, like

CSEDU 2018 - 10th International Conference on Computer Supported Education

260

Fulldome (Lantz, 2007). In this immersive environ-

ment, the projection occurs through different projec-

tors, some of which are polarized to the right eye and

others to the left eye. Thus, stereoscopic images are

formed with the aid of glasses. In addition, there is the

possibility of interaction with controls present in each

chair, so that the teacher can launch questions or chal-

lenges and the student responds directly my means of

these controls. Figure 6 shows the Immersive Theater

inside

Figure 6: Inside the Immersive Theater.

The aim of this article is to present examples of

LOs generated by F2S2 and the application possibil-

ities in different knowledge areas. So the remaining

part of the article is divided into three sections: in sec-

tion 2, the methodology, in section 3, the applications,

and in section 4, the conclusions.

2 METHODS

Three phases are stipulated for the generation of ob-

jects: acquisition, processing, and visualization. In

the acquisition process, the F2S2 is prepared, choos-

ing the background color, setting the camera, setting

the number of Frames and Streams, and lighting. In

the processing stage, lens correction, photo context

reduction, segmentation, and the organization of the

Streams are accomplished. Finally, the visualization

stage consists of adapting the Streams to the desired

projection formats and the conﬁguration of the nav-

igation software. Next, these three phases are ex-

plained in more detail, starting with the acquisition.

2.1 Aquisition

The F2S2 is controlled by a RaspBerry PI

and an

Arduino

, in addition to other electronic components

and motors. A communication protocol has been de-

veloped that transmits instructions for moving and

controlling the camera. The protocol is loaded into

these controllers. The object to be scanned is placed

Screenshot of Digital Arena

http://www.ftddigitalarena.com.br/wp-content/360tour/

index.html

https://www.raspberrypi.org/

https://www.arduino.cc/

in an inﬁnite background that may be covered with

different materials and colors in order to facilitate seg-

mentation (as described in section 2.2). The camera

and lighting settings are set manually as well as the

background color, which depends on the object. This

is the only manual step in the entire process.

Acquisitions occur by moving a camera over an

arc-shaped path based on the size of the object. Figure

7 shows the semi-sphere formed by the arc during the

scanning process.

Figure 7: Semi-spherical movement.

The angular variation θ represents the object’s ro-

tation over its own axis and ϕ represents the camera’s

angle relative to the object. The variable d represents

the distance between the camera and the center of the

object, that is, the radius of the semi-sphere, which is

constant throughout the movement.

To perform the movement, the F2S2 uses four

degrees of freedom, of which two are the X and

Y − axes, responsible for the arc displacement of the

camera. These axes start from an initial position de-

termined as angle zero (ϕ), traversing a path until

reaching its endpoint: the angle of ninety degrees rel-

ative to the base of the object.

The X − axis is responsible for the vertical move-

ment of the arm that supports the camera. The Y −

axis horizontally moves the arm of the X −axis, keep-

ing the camera in the arc motion at a distance d from

the object. The other two degrees of freedom are the

Z and B − axes, represented respectively by ϕ and θ

in Figure 7. Figure 8 shows these axes.

(a) (b)

Figure 8: Axes: (a) X and Y-axes, (b) Z and B-axes.

The image acquisition process starts with all axes

set to the zero-angle position. The object sits on a

turntable in the center of the B−axis. The acquisition

takes place through up to 360 shots determined by the

angle (θ) of the turn of the dish, that is, one can take a

Generation of Stereoscopic Interactive Learning Objects True to the Original Object

261

photo for each degree. The set of photos for the same

camera position is called Stream and each photo is

called Frame.

When the Stream is acquired at a certain angle,

the camera is repositioned one degree up in the arc,

i.e., the X and Y − axes are moved. In addition, to

keep the object in the focal center of the camera, the

Z − axis is moved to an angle determined by ϕ. The

maximum of Streams that can be acquired is ninety,

i.e., a Stream for each degree in the arc. Hence, it is

possible to acquire 360 Frames per Stream and ninety

Streams, giving a total of 32,400 photos, represent-

ing each possible view degree by degree of the object.

Figure 9 shows two views of F2S2.

(a) (b)

Figure 9: Views of F2S2: (a) lateral view, (b) internal view.

It takes about 30 hours to complete the acquisition

of an object.

2.2 Processing

Camera lenses cause deformation in the acquired im-

ages, which is inherent to all lenses and cameras. De-

spite the little importance of it in amateur photos, in

professional/scientiﬁc processes, it requires correc-

tion. There is specialized software that can perform

the process automatically on a set of images from

a conﬁguration chosen by the user. An example of

free software that accomplishes this job is Darktable

Thus, processing starts with correcting deformities in

all Frames.

Depending on the size and shape of the object,

it may be impossible to frame it so that it occupies

the entire space in the picture. To reduce the amount

of space required for disk storage and eliminate un-

wanted parts, such as parts of the scanner that may

appear in the photos, context reduction is applied by

selecting a region where the object appears integrally

in all Frames. The rest of the image is discarded. Fig-

ure 10(a) shows a complete image, and Figure 10(b)

shows the same image with context reduction.

The segmentation step (Pedro et al., 2013;

Chauhan et al., 2014)removes the background of the

www.darktable.org

(a) (b)

Figure 10: Context Reduction: (a) original image, (b) image

with context reduction.

image, keeping only the object of interest. There is

no ideal solution in the literature for the segmentation

problem since each image has its own characteristics.

In addition, the acquisition generates a high volume

of images. Hence, the process has to be automated.

In order to bypass these problems, we have

tested several algorithms, such as Support Vector

Machine (SVM)(Huang et al., 2012; Tsai et al.,

2006), K-means (Xu et al., 2017), K-Nearest Neigh-

bor (KNN)(Xu and Wunsch, 2005), Sobel (Jo and

Lee, 2012), Canny (Chen et al., 2014), Graph Cut

(Boykov and Jolly, 2001), and Watershed (Tian and

Yu, 2016). The algorithm with the best results, con-

sidering processing time and quality, was the SVM.

Figure 11 shows the result of a segmentation per-

formed on a Darth Vader doll on a blue background.

(a) (b)

Figure 11: Segmentation: (a) Image with context reduction,

(b) Segmented image.

The Streams are organized during the segmenta-

tion of the images when they receive a ﬁlename con-

taining the identiﬁcation of the Stream to which they

belong and the Frame; for instance, the ﬁrst image of

the ﬁrst Stream is named S001F0001.

The equipment used for the job was an Avell Full-

range W1713 Pro CL notebook with an Intel Core I7-

4910MQ 2.9GHz processor, 32GB RAM, NVIDIA

Quadro K3100M and Intel HD Graphics 4600 graph-

ics card, and Windows 10 PRO 64bits operating sys-

tem. The Streams context reduction, segmentation,

and organization system was developed using Matlab

CSEDU 2018 - 10th International Conference on Computer Supported Education

262

Toolbox

v.R2015b, and OpenCV-python 3.0 v.3.6.2

of Python

The processing time for a set of 32,400 images

was 12 hours and 30 minutes, occupying 1.2 Giga-

bytes of memory, which is acceptable considering

that the acquisition time is longer than the processing

time.

2.3 Visualization

As each device has its speciﬁcity, it is necessary to

prepare the images for each of them. Stereoscopic

devices have their own format for visualization, such

as anaglyph, polarized and stereo pair, among others

(Oh et al., 2017; Liu et al., 2017; Lei et al., 2017).

The anaglyph consists of two superimposed images of

the same object with different color patterns - usually

red and blue or red and green - each pattern having a

small displacement with respect to one another. Spe-

cial eyeglasses are used to ﬁlter each pattern for one

eye. The notion of depth is generated by the differ-

ences in patterns and the small displacement.

Polarized stereoscopy consists of two overlapping

images, but it requires the use of a pair of glasses in

which the lenses are prepared to ﬁlter the polarized

light so that each eye receives each of the images.

The stereoscopic pair consists of two images, one of

which is horizontally polarized and the other verti-

cally. Thus, the device emits each image in a different

polarization but simultaneously. When the glasses re-

ceive these images, the lenses ﬁlter each one for each

eye, causing the brain to interpret the image with a

notion of depth.

F2S2 is able to acquire stereoscopic images using

stereoscopic lenses developed by the manufacturers

of the cameras used, but it is also capable of generat-

ing stereoscopy through lenses without this function-

ality. For the latter, the formation of stereoscopic vi-

sion occurs by joining the acquired images at different

angles, for example, one image at angle zero and an-

other at the angle of 4 degrees from zero. This value

is related to the distance between the camera and the

object, which, in turn, is related to Parallax and Dis-

parity, as can be seen in (Kramida, 2016).

The Stereo Pair, on the other hand, is based on two

images, each of which is displayed for only one eye,

which has the same effect as polarized stereoscopy.

The technique of the Stereo Pair, which has been

known since the mid-nineteenth century, consists of

Matlab Imaging Processing Toolbox:

https://www.mathworks.com/products/image.html?s tid=

srchtitle

OpenCV-Python: https://pypi.python.org/pypi/opencv-

python

placing side by side two images acquired at different

angles, as cited above.

In the case of the images generated by the F2S2,

they require speciﬁc processing to transform them

into stereoscopic images. This processing is the junc-

tion of the images of certain angles related to the par-

allax and disparity, the average distance between the

retinas, which varies from 4 to 7 centimeter.

As far as navigation is concerned, two ways have

been developed: if the device has enough memory

to support all images, they are loaded into memory.

The change in visualization takes place according to

the interaction of the user. If he presses the left or

right arrow key, for example, the picture changes to

the Frame immediately to the right or left of the cur-

rent frame. This change of Frames is related to the

change in the angles of axis B. Changes up or down,

using the respective arrow keys, imply changes in the

Streams, which relates to the angles in the arc as de-

ﬁned in subsection 2.1.

The same process occurs when using VR glasses,

but instead of pressing keys, it reads the accelerome-

ter and gyro signals to carry out the change of Frames

or Streams. In addition, joysticks may be used, which

allow interaction by pressing their buttons.

The interactions also allow zooming in and out,

giving the sensation of approaching or distancing

from the object, and translation, i.e., the displacement

of the object in the projection.

However, if there is not enough memory to load

all the images, navigation takes place through a mov-

ing windows. It works by loading a subset of the

images closest to the current view. When moving

within the window, which means to change Frame or

Stream, the closest images are found, unloading those

Frames that, according to the new position, have be-

come more distant and loading the new closer ones.

The size of the windows, that is, the number

of Frames loaded, is conﬁgurable depending on the

equipment (memory and processor) used. Not all

equipment will be able to run the browser, given the

size and quantity of images. Navigation itself occurs

in a similar way to that described in the ﬁrst case.

3 APPLICATIONS

In this section, we present some possibilities of using

the LOs in the Immersive Theater (Fulldome), where

we used human skulls to study anatomy, the Pyramid,

where we used a bronze bust to study the arts, Vir-

tual Reality Glasses, using the inside of a car (starter

motor) for studies in the area of mechanics, and the

TV / Projector, where we used a replica of a car for

Generation of Stereoscopic Interactive Learning Objects True to the Original Object

263

the study in design. This section has been divided

into subsections that address each environment with

its LO

3.1 Immersive Theater - LO Skulls

The immersive theater used has four polarized projec-

tors to generate stereoscopy, with each pair projecting

an image for each eye. The chairs are equipped with

controls that allow interaction. The exhibition room

has 120 seats and seats for wheelchair users. As the

projection screen format is semi-spherical and has a

diameter of 14 meters, the viewers should be slightly

inclined to have a full view.

Two human skulls, one with the skull box open

and the other with it closed, were made available by

the medical course. They were digitalized obtaining

two LOs in UHD, which allows the visualization of

anatomical details. The LOs present ﬁdelity of shape,

color, and texture to the original pieces. There are,

of course, images of skulls created in 3D drawing

and modeling programs to which a texturing has been

applied; this is called rendering. In comparison to

the O2 generated by F2S2, however, these rendered

pieces do not have the three ﬁdelity characteristics,

which can be seen in Figure 12.

(a) (b)

Figure 12: Comparison between Skulls: (a) Skull generated

by F2S2, (b) Skull generated by 123D, (c) Skull Generated

by F2S2, and (d) Synthetic Skull.

Figures 12(a) and 12(b) show the difference be-

tween an image generated by F2S2 and one generated

by the AutoDesk 123D Catch

software. The latter

We present the videos of the LOs in project’s

channel in youtube https://www.youtube.com/channel/

UCXEFzyZGrGlLANNrCawNgWQ - Anonymous.

https://www.autodesk.com.br/

presents ﬁdelity problems in the rendering process,

which does not occur in the LO generated by F2S2.

Figures 14 (c) and 14 (d) show the difference between

the LO generated by F2S2 and a synthetic 3D model.

As can be seen in Figure 12(d), the synthetic model

does not represent the ear hole.

The skull with the open skull box was used for

the generation of an LO for the immersive theater,

with the main objective being the anatomical study in

the medical area; however, the same LO may be used

in different courses (Dragon et al., 2013) such as an-

thropology, physiotherapy, nursing, and even in high

school. The main advantage, in courses that do not

require manipulating corpses in laboratories, which is

unhealthy, is that the ﬁdelity allows a detailed visual-

ization of the piece. Figure 13 shows the preparation

of the LO for immersive theater projection.

Figure 13: LO immersive theater preparation.

3.2 Pyramid - Bronze Bust

The pyramid used has three faces, made of glass for

a better reﬂection of the object. It is 1.05 meters at

the base, 48 cm high and 67 cm deep. For this ex-

periment, we used a bronze bust. The purpose of this

LO is its use in areas such as arts, history and even to

create a virtual museum. This LO allows the study of

bronze sculpture techniques, the analysis of the char-

acteristics of the piece (art history), and even the study

of the character represented in the piece (history).

This LO projected on the pyramid conveys the

feeling that the piece is physically inside it, but it is

only a visual representation. There is also the feel-

ing of three-dimensionality. The teacher can place

points of interest that he can access through keyboard,

mouse or joystick interaction. Figure 14 shows the

pyramid with the bronze bust.

3.3 VR Glasses - Starter Motor

Another relevant purpose for F2S2 is the generation

of LOs for inspection. In this context, a starter mo-

tor, a piece that starts the rotation process of the car’s

engine, was used for the LO. The inspection process

can also be performed on other mechanical pieces or

on electronic equipment and components.

CSEDU 2018 - 10th International Conference on Computer Supported Education

264

Figure 14: LO Pyramid preparation.

We used RV spectacles for the presentation of this

LO. The characteristic of these glasses is that they are

intrinsically stereoscopic, since there is a division in

the device itself and, consequently, it separates an im-

age for each eye. The advantage is that, as it uses the

accelerometer and gyroscope, the visualization of the

LO is natural, that is, it does not require keyboard and

mouse for interaction. One is free to move around in

the environment without losing the focus of the LO.

A disadvantage is the reduction of the quality and

quantity of images, since cell phones, for example,

have limitations in computational capacity. Although

there are ﬁxed-screen glasses with a cable connec-

tion to a computer, this disadvantages the user’s free

movement, limited by the length of the cable. Figure

15 shows the use of LO in RV glasses.

Figure 15: LO on VR Glasses.

3.4 TV/Projector - Miniature Replica of

a Car

The TV/projector environment is the one that has

greater availability of use since schools and univer-

sities have easy access. For this experiment, a replica

of a Corvette car was chosen, aiming at studying its

design, which may be of interest to industrial design

and design courses, for instance.

The advantage of the LO on TVs is that TVs may

have a UHD resolution of 4K and 8K, allowing a

detailed view of the images provided as the images

themselves also have a 4K and 8K deﬁnition. An-

other feature of the TV is that it can be stereoscopic,

using the technique of polarized stereoscopy. Figure

16 shows the LO displayed on a 4K TV.

Figure 16: LO on a 4K TV.

4 CONCLUSIONS

The use of F2S2 in educational environments insti-

gates multidisciplinarity and interdisciplinarity since

the same Learning Object may be reused in different

areas of knowledge and addressed in different ways in

the same area.

An example of this is the Bronze Bust, which may

be used in History for the study of the character or the

analysis of the artistic movement at the time in which

the piece was created. It may also serve different dis-

ciplines, such as those in the arts course.

Thus, we have shown the feasibility of using the

learning objects generated by F2S2, even as a way to

promote interaction and dynamism in educational en-

vironments. This is mainly due to the approximation

of technologies that are currently present in areas such

as ﬁlm and advertising. We believe that these tech-

nologies may arouse interest, since people also have

contact with some of these technologies in their daily

lives, outside the school environment.

The comprehensiveness of the learning objects

generated by F2S2 goes beyond disciplinary bound-

aries and knowledge areas since the same object may

be used in both the human and the exact sciences.

In the future, we plan to accomplish the scanning

of other objects and make the object bases as well as

the visualization software available. We also aim to

evaluate the learning objects with students, in order

to verify interaction and aroused interest.

ACKNOWLEDGEMENTS

We gratefully acknowledge the support of IFPR,

UTFPR, CAPES and Nvidia GPU Grant Program for

the donation of the TITAN XP GPU.

Generation of Stereoscopic Interactive Learning Objects True to the Original Object

265

REFERENCES

Boykov, Y. Y. and Jolly, M. P. (2001). Interactive graph

cuts for optimal boundary amp; region segmentation

of objects in n-d images. In Proceedings Eighth IEEE

International Conference on Computer Vision. ICCV

2001, volume 1, pages 105–112 vol.1.

Chauhan, A. S., Silakari, S., and Dixit, M. (2014). Image

segmentation methods: A survey approach. In 2014

Fourth International Conference on Communication

Systems and Network Technologies, pages 929–933.

Chen, H., Ding, H., He, X., and Zhuang, H. (2014). Color

image segmentation based on seeded region growing

with canny edge detection. In 12th International Con-

ference on Signal Processing (ICSP), pages 683–686.

Dragon, T., Mavrikis, M., McLaren, B. M., Harrer, A.,

Kynigos, C., Wegerif, R., and Yang, Y. (2013).

Metafora: A web-based platform for learning to learn

together in science and mathematics. IEEE Transac-

tions on Learning Technologies, 6(3):197–207.

Freire, M. and Fern

andez-Manj

on, B. (2016). Metadata for

serious games in learning object repositories. IEEE

Revista Iberoamericana de Tecnologias del Apren-

dizaje, 11(2):95–100.

Garrido, A. and Onaindia, E. (2013). Assembling learning

objects for personalized learning: An ai planning per-

spective. IEEE Intelligent Systems, 28(2):64–73.

Hassan, H., Martinez-Rubio, J. M., Perles, A., Capella,

J. V., Dominguez, C., and Albaladejo, J. (2013).

Smartphone-based industrial informatics projects and

laboratories. IEEE Transactions on Industrial Infor-

matics, 9(1):557–566.

Huang, G. B., Zhou, H., Ding, X., and Zhang, R. (2012).

Extreme learning machine for regression and mul-

ticlass classiﬁcation. IEEE Transactions on Sys-

tems, Man, and Cybernetics, Part B (Cybernetics),

42(2):513–529.

Hung, I.-C., Kinshuk, and Chen, N.-S. (2017). Embodied

interactive video lectures for improving learning com-

prehension and retention. Computers and Education.

Jo, J. H. and Lee, S. G. (2012). Sobel mask operations

using shared memory in cuda environment. In 2012

6th International Conference on New Trends in In-

formation Science, Service Science and Data Mining

(ISSDM2012), pages 289–292.

Koukopoulos, Z. and Koukopoulos, D. (2017). Integrat-

ing educational theories into a feasible digital envi-

ronment. Applied Computing and Informatics.

Kramida, G. (2016). Resolving the vergence-

accommodation conﬂict in head-mounted displays.

IEEE Transactions on Visualization and Computer

Graphics, 22(7):1912–1931.

Lantz, E. (2007). A survey of large-scale immersive dis-

plays. In Proceedings of the 2007 Workshop on

Emerging Displays Technologies: Images and Be-

yond: The Future of Displays and Interacton, EDT

’07, New York, NY, USA. ACM.

Lei, J., Wu, M., Zhang, C., Wu, F., Ling, N., and Hou,

C. (2017). Depth-preserving stereo image retargeting

based on pixel fusion. IEEE Transactions on Multi-

media, 19(7):1442–1453.

Liu, Z., Yang, C., Rho, S., Liu, S., and Jiang, F. (2017).

Structured entropy of primitive: big data-based stereo-

scopic image quality assessment. IET Image Process-

ing, 11(10):854–860.

endez, N. D. D., Morales, V. T., and Vicari, R. M.

(2016). Learning object metadata mapping with

learning styles as a strategy for improving usabil-

ity of educational resource repositories. IEEE Re-

vista Iberoamericana de Tecnologias del Aprendizaje,

11(2):101–106.

Oh, H., Kim, J., Kim, J., Kim, T., Lee, S., and Bovik, A. C.

(2017). Enhancement of visual comfort and sense of

presence on stereoscopic 3d images. IEEE Transac-

tions on Image Processing, 26(8):3789–3801.

Pedro, R. W. D., Nunes, F. L. S., and Machado-Lima, A.

(2013). Using grammars for pattern recognition in

images: A systematic review. ACM Comput. Surv.,

46(2):26:1–26:34.

Rossi, D. (2015). A hand-held 3d-printed box projector

study for a souvenir from a mixed-reality experience.

In 2015 Digital Heritage, volume 1, pages 313–316.

Santiago, R. and Raabe, A. (2010). Architecture for learn-

ing objects sharing among learning institutions 2014

lop2p. IEEE Transactions on Learning Technologies,

3(2):91–95.

Sanz-Rodriguez, J., Dodero, J. M. M., and Sanchez-Alonso,

S. (2010). Ranking learning objects through integra-

tion of different quality indicators. IEEE Transactions

on Learning Technologies, 3(4):358–363.

Sidharta, R., Hiyama, A., Tanikawa, T., and Hirose, M.

(2007). Volumetric display for augmented reality.

In 17th International Conference on Artiﬁcial Reality

and Telexistence (ICAT 2007), pages 55–62.

Sinclair, J., Joy, M., Yau, J. Y.-K., and Hagan, S. (2013).

A practice-oriented review of learning objects. IEEE

Transactions on Learning Technologies, 6(2):177–

192.

Tian, X. and Yu, W. (2016). Color image segmentation

based on watershed transform and feature clustering.

In 2016 IEEE Advanced Information Management,

Communicates, Electronic and Automation Control

Conference (IMCEC), pages 1830–1833.

Tsai, C.-F., McGarry, K., and Tait, J. (2006). Claire: A mod-

ular support vector image indexing and classiﬁcation

system. ACM Trans. Inf. Syst., 24(3):353–379.

Wiley, D. A. (2002). The instructional use of learn-

ing objects: Online version. http://www.test.org/doe/

acessed in 2017/10/15.

Xu, J., Han, J., Nie, F., and Li, X. (2017). Re-weighted

discriminatively embedded k -means for multi-view

clustering. IEEE Transactions on Image Processing,

26(6):3016–3027.

Xu, R. and Wunsch, D. (2005). Survey of clustering al-

gorithms. IEEE Transactions on Neural Networks,

16(3):645–678.

CSEDU 2018 - 10th International Conference on Computer Supported Education

266